In single-plate color solid-state imaging devices as typified by CMOS image sensors, three or tour kinds of color filters are arranged in mosaic form on a large number of photoelectric conversion pixels that are arranged on a semiconductor substrate, With this structure, color signals corresponding to the color filters are output from the pixels, respectively, and a color image is generated by performing signal processing on those color signals.
However, color solid-state imaging devices in which color filters of the primary colors are arranged in mosaic form have a problem that they are low in efficiency of light utilization and sensitivity because ⅔ of incident light is absorbed by the color filters. The fact that each pixel produces a color signal of only one color raises a problem of low resolution. In particular, false colors appear noticeably.
To solve the above problems, imaging devices having a structure that photoelectric conversion layers are laminated in three layers on a semiconductor substrate on which signal reading circuits are formed are being studied and developed (refer to JP-T-2002-502120 (The symbol “JP-T” as used herein means a published Japanese translation of a PCT patent application.) (corresponding to WO 99/39372) and JP-A-2002-83946, for example). For example, these imaging devices have a pixel structure that photoelectric conversion layers which generate signal charges (electrons or holes) in response to blue (B) light, green (G) light, and red (R) light are laid in this order from the light incidence surface. Furthermore, these imaging devices are provided with signal reading circuits capable of independently reading, on a pixel-by-pixel basis, signal charges generated by the photoelectric conversion layers.
In the imaging devices having the above structure, almost all of incident light is photoelectrically converted into signal charges to be read and hence the efficiency of utilization of visible light is close to 100%. Furthermore, since respective pixels produce color signals of the three colors (R, G, and B), these imaging devices can generate good, high-resolution images (no false colors appear noticeably) with high sensitivity.
In the imaging device disclosed in JP-T-2002-513145 (corresponding to WO 99/56097), triple wells (photodiodes) for detecting optical signals are formed in a silicon substrate and signals having different spectra (i.e., having peaks at B (blue), G (green), and R (red) wavelengths in this order from the surface) are obtained so as to correspond to different depths in the silicon substrate. This utilizes the fact that the distance of entrance of incident light into the silicon substrate depends on the wavelength. Like the imaging devices disclosed in JP-T-2002-502120 (corresponding to WO 99/39372) and JP-A-2002-83946, this imaging device can produce good, high-resolution images (no false colors appear noticeably) with high sensitivity.
However, in the imaging devices disclosed in JP-T-2002-502120 (corresponding to WO 99/39372) and JP-A-2002-83946, it is necessary that the photoelectric conversion layers be formed in order in three layers on the semiconductor substrate and vertical interconnections be formed which transmit R, G, and B signal charges generated in the respective photoelectric conversion layers to the signal reading circuits formed on the semiconductor substrate. As such, these imaging devices have problems that they are difficult to manufacture and they are costly because of low production yields.
On the other hand, the imaging device disclosed in JP-T-2002-513145 corresponding to WO 99/56097) is configured in such a manner that blue light is detected by the shallowest photodiodes, red light is detected by the deepest photodiodes, and green light is detected by the intermediate photodiodes. However, the shallowest photodiodes, for example, also generate photo charges when receiving green or red light, as a result of which the spectra of R, G, an dB signals are not separated sufficiently from each other. Therefore, to obtain true R, G, and B signals, it is necessary to perform addition/subtraction processing on output signals of the photodiodes, which means a heavy computation load. Another problem is that the addition/subtraction processing lowers the S/N ratios of the image signals.
The imaging device disclosed in JP-A-2003-332551 has been proposed as one capable of solving the problems of the imaging devices of JP-T-2002-502120 (corresponding to WO 99/39372), JP-A-2002-83946 and JP-T-2002-513145 (corresponding to WO 99/56097). This imaging device is a hybrid type of the imaging devices of JP-T-2002-502120 (corresponding to WO 99/39372) and JP-A-2002-83946 and the imaging device of JP-T-2002513145 (corresponding to WO 99/56097) and is configured as follows. Only a photoelectric conversion layer (one layer) that is sensitive to green (G) light is laid on a semiconductor substrate and, as in the previous image sensors, incident light of blue (B) and red (R) that has passed through the photoelectric conversion layer is detected by two sets of photodiodes that are formed in the semiconductor substrate so as to be arranged in its depth direction.
Since it is sufficient to form only one photoelectric conversion layer, the manufacturing process is simplified and cost increase or reduction in yield can be avoided. Furthermore, since green light which is in an intermediate wavelength range is absorbed by the photoelectric conversion layer, the separation between the spectrum of signals generated by the photodiodes for blue light formed in the semiconductor substrate and that generated by the photodiodes for red light is improved, whereby the color reproduction performance is improved and the S/N ratios are increased.